The Altair Community is migrating to a new platform to provide a better experience for you. In preparation for the migration, the Altair Community is on read-only mode from October 28 - November 6, 2024. Technical support via cases will continue to work as is. For any urgent requests from Students/Faculty members, please submit the form linked here

"More neurons in hidden layer wont increase/decrease neural network performance"

YvesAYvesA Member Posts: 1 Learner III
edited June 2019 in Help
Hi everyone,

i used the search function but found nothing.

I'm using RapidMiner 5.1.003

Data Mining Setup is as following:

i've used 7 different datasets from the uci repository (wine,adult,krkrpa7,iris,house-votes-84, heart-cleveland and zoo) for a classification task. In all of those i meassured performance using cross-validation (10 validations, shuffled or linear sampling)

In all of those i was wondering why an increase in hidden layer neurons wont have any impact on the result. With more hidden neurons it should perform better unless it's overfitting. Am i wrong somewhere?Is this just concerning the free community edition?

The neural network model in average works fine, i was just wondering why there is no slight increase/decrease in performance.

thanks for reading & have a nice day
yves

Answers

  • Nils_WoehlerNils_Woehler Member Posts: 463 Maven
    Hi,

    what settings did you choose for your hidden layers? I just tested it with the sonar dataset from the sample directory of RapidMiner and for me it works fine.
    With just one layer the performance is at 79%. With two layers it increases to 83% and with four layers it drops to 50%.

    Best,
    Nils
Sign In or Register to comment.